Working Paper No. 257, April, Abhijit V. Banerjee and Sendhil Mullainathan (2010)

This paper constructs a model of temptation goods, showing that spending on them is more consequential for the poor than for the rich, and with this model explains many of the anomalies observed in the spending patterns of the poor.

Many scholars and development workers have observed that in their investment decisions, the poor “often behave as if they are very myopic.”  Their spending patterns frequently value consumption today over consumption tomorrow.  Scholars have developed several theories about why this might be the case.  Firstly, the poor might face a minimum consumption constraint.  Secondly, they might choose consumption today because of high mortality rates.  Thirdly, they might choose to pay high interest rates to finance consumption today if in reality they do not end up paying them because of default.

However, none of these theories square with the empirical evidence.  The poor spend a large fraction of their income on non-essential goods, such as cigarettes, alcohol, and non-nutritious foods, and their spending exhibits high frequency volatility.  Adult mortality rates are not substantially different for the poor and non-poor.  And default rates among the poor are extremely low. 

If none of these theories hold water, why do the poor empirically value consumption today more than the rich?  Banerjee and Mullainathan propose an interesting theoretical model to explain this difference in spending patterns.  They assume that there are two types of goods—those that generate utility during consumption and before consumption and those that generate utility only during consumption.  They explain that a donut or cigarette “represents a temptation good.  In the moment we would spend money on it.  But we would like future selves to not spend money on it.  …This is in alternative to other goods where present and future selves agree: in the moment we spend on them and we would like future selves to spend on them as well.  These are goods like a good education for our child, or a nice house to retire in.” 

The observation upon which the rest of the model rests is that of “declining temptation,” or more specifically, that the impact of temptation goods is declining in income.  The authors assume that temptation goods—like a 25 cent donut—tempt poor and rich equally.  However, such an expenditure would have very different ramifications for someone earning $2 a day versus $30.  Through a series of theoretical models they show that the declining impact of temptation in income has wide ranging impacts for the economic lives of the poor, such as in savings, credit, risk taking, investment, credit availability, and poverty traps.

The authors envision a multi-period self in which today’s self “values his own spending on temptation goods but does not value tomorrow’s self’s spending on them.”  Moreover, the self knows that he will value temptation goods more when confronted with them than at any other time.  This is the “self-control problem.”  There are several implications of this model.

• Since the self knows about his or her self-control problem, he or she will be willing to pay extra for illiquid durables because they represent a commitment device in a situation in which income is taxed by temptation.

• Self-control problems will lead to demand for ways to commit to saving.  Empirical evidence supports this theory.  Ashraf, Karlan, and Yin (2004) find that even if they already have bank accounts, the poor will subscribe to a second account in which they can deposit money but not withdraw it until a certain date or level.

• Expectations of future wealth matter.  The standard economic savings model holds that an increase in future income will reduce savings today and increase consumption today.  In this way, consumption is smoothed over one’s lifetime.  However, if declining temptation is added to the mix, behaviors change.  The expectation of increased future income may encourage the poor to save more now.  And more seriously, the expectation of decreasing future income may cause the poor to consume more today.

• Standard economic theory holds that individuals will undertake investments with a positive net return.  However, under declining temptations, individuals may not invest in high return projects when the scale is too small.  The authors write that “This captures the lay intuition that investments may be unimportant unless they significantly change one’s circumstances.”  This explains several puzzles in current development literature.  For example, studies have found that some farmers do not buy fertilizer even when there would be high rates of return to doing so.  The authors reason that “Farmers might prefer to consume everything they have, because given the credit constraints, any interesting projects are beyond their reach.”

• Awareness of the self-control problem will alter the demand for credit.  Standard theory holds that credit is good because it expands the opportunity set, yet under declining temptations credit can be harmful.  This explains empirical observations such as that the poor demand microfinance yet avoid credit cards.

• Declining temptations also affects the behavior of lenders to the poor.  Under this model, there exist situations in which lenders would choose to block loans to poor borrowers for high return projects, because such projects would cause reduced borrowing.  Lenders may therefore prefer to keep borrowers in a poverty trap.

Overall, this study informs a literature dating back to 1932 when Irving Fisher observed that the poor tend to more myopic in their spending than the rich.  The authors write that “This is a direct consequence of declining temptations, which, it is worth emphasizing, does not say that the poor are more tempted.  Indeed the amount spent on temptation goods over the lifetime is strictly increasing in initial wealth; it is just that it takes a smaller fraction of total consumption to satisfy one’s temptations at high levels of consumption, which is why the same failures (giving in to temptations) have greater consequences when poor.”  Understanding temptations is important because as shown, models that take them into account (and in particular, the fact that giving into them is more consequential for the poor than the rich) predict significantly different behaviour than standard economic models—from borrowing preferences to consumption decisions to the credit available to the poor.